VR2R-0694_r1

VR2R-0694

① SA Source

• Source: 開啟完整 SA 文章
• Section: VR NVL72 TCO: BoM and Power Budget Analysis
• Line hint: 601

Context Before

The VR NVL72 is more expensive on a per-GPU capital cost basis, ~45% higher vs GB300s and ~14-15% higher vs the MI4XX given a higher server cost on a per GPU basis. This results in a higher Capital Cost of Ownership (TCO).

Evidence

For example VR NVL 72 Hyperscaler Arista has a capital cost of $3.28to[[04_{k}nowledg{e}_{b}ase/MI400|MI4XX]][[04_{k}nowledg{e}_{b}ase/Hyperscaler|Hyperscaler]]of$2.86 per hour per GPU over a 4 year useful life.

Context After

Our TCO Model runs on a 4y useful life for the purpose of calculating capital cost per hour to reflect a conservative business case, but most Neoclouds and Hyperscalers will use a 5-6y depreciation period and we think it is best to look at EBIT margins using this depreciation period. Our preferred yardstick is Project IRR, which is agnostic to the chosen depreciation period.

② Atomic Claim

例如，以 4 年 useful life 計算，採用 Arista networking 的 VR NVL72 Hyperscaler 部署，其 capital cost 為每顆 GPU 每小時 3.28 美元；相較之下，MI4XX Hyperscaler 為 2.86 美元。

• Epistemic Mode: ASSERTED
• Mapping Status: COMPLETE

③ Semantic Frame

{
  "frame_type": "NARY_RELATION",
  "participants": [
    {
      "node": {
        "id": "04_knowledge_base/VR NVL72",
        "label": "VR NVL72"
      },
      "role": "subject_platform"
    },
    {
      "node": {
        "id": "04_knowledge_base/MI400",
        "label": "MI4XX"
      },
      "role": "comparison_platform"
    },
    {
      "node": {
        "id": "04_knowledge_base/Hyperscaler",
        "label": "Hyperscaler"
      },
      "role": "deployment_type"
    },
    {
      "node": {
        "id": "02_companies/Arista",
        "label": "Arista"
      },
      "role": "network_vendor"
    },
    {
      "node": {
        "id": "04_knowledge_base/GPU",
        "label": "GPU"
      },
      "role": "cost_unit"
    }
  ],
  "qualifiers": {
    "condition_text": "capital cost per hour per GPU",
    "numeric_mentions": [
      "$3.28",
      "$2.86",
      "4 year"
    ],
    "temporal_mentions": [
      "4 year useful life"
    ]
  },
  "relation_type": "VR_NVL72_VS_MI4XX_CAPITAL_COST_PER_GPU_HOUR"
}

④ Canonical Entity Mapping

Role	Surface Label	Canonical Target
subject_platform	VR NVL72	04_knowledge_base/VR NVL72
comparison_platform	MI4XX	MI400
deployment_type	Hyperscaler	Hyperscaler
network_vendor	Arista	Arista
cost_unit	GPU	GPU

⑤ Human Review

請在 Properties 逐項確認：

• 原文 → Atomic Claim 是否忠實
• Atomic Claim → Semantic Frame 是否忠實
• Canonical Entity mapping 是否正確
• Epistemic mode 是否保留原文語氣
• 最後選擇 review_action

Review state

Markdown 內文不是正式 approval。只有 Apply bridge 寫入的 Decision Ledger event 才是正式決策。